The project ended in August 2019 and released quite the number of significant results and innovations:
Through a constant interaction with stakeholders in the field of environmental monitoring, user needs and requirements for an efficient citizen observatory on land resource monitoring were elicited, since the early steps of the project. Following this, Scent toolbox system requirements and architecture were defined aiming to respond to these needs and to allow the deployment of the toolbox to the pilot areas.
SCENT crowdsourcing and serious gaming applications have been implemented aiming to enable, incentivise and increase the involvement of citizens as individuals and/or citizen groups/associations in the process of environmental monitoring. Through SCENT Explore, citizens can collect images and videos of LC/LU elements and other environmental parameters (water level and flow velocity) along with textual descriptions, and report flood related events like flooded locations, etc. SCENT Collaborate and CAPTCHA plugin have been also implemented, enabling users to validate the textual description of the collected images simply by browsing images and stating whether for example, they can see forests, river banks, buildings, etc. Moreover, SCENT Measure application has been implemented, allowing the communication with low-cost portable sensors and the collection and visualisation of soil moisture and air temperature observations.
The design and development of SCENT crowdsourcing platform was realised, acting as a key data exchange component for all citizen-generated data, whilst also connecting the SCENT crowdsourcing frontends (SCENT Explore, Measure, Collaborate & Captcha) with other core parts of the toolbox (i.e. Scent Intelligence Engine and with the Harmonisation platform). The platform implements a data qualification mechanism in charge of ensuring that annotations associated to images are of sufficient quality for being further used in decision making and flood modeling as well as a crawling functionality that can be used to fetch relevant images from open data platforms. Moreover, a web-based application- the SCENT Campaign Manager has been developed, enabling relevant authorities (i.e. policy makers, etc) to design and manage citizen-science campaigns targeted at specific areas where relevant environmental information is needed.
The Image Analysis Tool of the Scent Intelligence Engine has been designed and implemented, being capable of analysing images collected from citizens and open platforms and classifying detected items through a specifically developed set of deep learning networks optimized for SCENT oriented taxonomy classification and class region detection. In addition, the Map Segmentation, Delineation and Annotation Tool (MSDAT) has been implemented, enabling the automated detection and annotation of segments on satellite or aerial maps with those LC/LU SCENT taxonomy classes that affect water flow the most. It returns a semantically meaningful raster map represented as a taxonomy layer on top of the acquired satellite maps.
In order to augment costly in-situ sensors and collect river measurements with a consistent and accurate way from volunteers, a water level measurement tool, capable to receive an image of a water level indicator that is half-submerged into water and extract the water level measurement, and a water velocity calculation tool, capable to receive a video of a pre-defined floating object in a water body and calculate the surface velocity of the water, have been developed.
A set of hydrodynamic and hydrological models of Danube Delta and Kifisos pilot cases have been developed using existing in-situ data. These models were properly configured to allow integration of citizen-science data collected through SCENT campaigns, whilst model results were assessed showcasing improvements in comparison with the initial models.
The SCENT Harmonisation platform has been implemented that is tasked with the storage, management and provision of all the data collected or created by the various SCENT components. To ensure the quality of the information received from the portable sensors, a Sensor Data Quality mechanism has been implemented, that applies different statistical methods and tests in order to cope with aspects related to the correctness/accuracy of the sensor measurements, user reliability, and protection of the system from malicious contributions. In addition, the Harmonisation platform ensures the interoperability of the citizen-science data (as well as of the rest of the datasets managed by the platform), enabling the translation of the collected data to Open Geospatial Consortium (OGC) compliant observations and the provision of this information either through interactive visualisations (dashboards) or through proper APIs. Last but not least, SCENT project has been established as a GEOSS Data Provider, making available its resources through GEOSS portal.
Various citizen-science campaigns (10 in total) were designed and conducted in Kifisos and Danube Delta pilot sites. These field campaigns were organised in several different dedicated periods in both pilots covering in total about 11 months, while focusing on themes of interest to the local communities and the policy makers such as the collection of LC/LU elements, river parameters and soil measurements. More than 700 volunteers participated in the field campaigns, testing, evaluating and validating the SCENT toolbox while also contributing with a significant amount of 24200+ environmental resources in both pilot areas. It should be noted, that SCENT has established an active online community involving, as of August 2019, 1130+ active registered users of SCENT Collaborate, while SCENT Explore and Measure downloads have reached 2000+, with 612 active users.
